Modelling and Identification of the Human Balancing System

Master Thesis presentation

Raphaëlle Peyraud


Thursday, September 12, 2019, 14:00 - 15:00

Room 01-012, Georges-Köhler-Allee 102, Freiburg 79110, Germany

Humans maintain stable balance easily in a lot of different environments, and against different external perturbations. Ageing or diseases can affect the control mechanism of the human upright stance, which results in a fail or difficulties of maintaining balance for the affected people. In order to better understand this mechanism, models have been developed, and parameters estimation based on experimental data has been carried out, because they represent the state of the system. Even if effort was put to keep those models simple, they still contain nonlinearities.

Until now, the optimization fitting was realized in the frequency domain, but it remains a challenge to implement the models. To overcome this limitation, the two provided models (Independent Channels, IC) and (Disturbance Estimation and Compensation, DEC) were formulated in a state-space formulation, and numerical optimization methods were implemented to estimate their parameter values.

The objective to minimize was derived following a maximum-likelihood estimation approach. A Nonlinear Program (NLP) was formulated, and the dynamics was discretized using the Multiple Shooting (MS) method. Moreover, because of the high dimensionality of the addressed problem, a suited integration method and sampling time were chosen.

A clean formulation of the models and of the optimization problem is now available, and the models' nonlinearities were successfully implemented. The identified parameters were able to fit the data and reproduced the sway response for both models.